Bugs etc.


Kernel message

Whoops: end_buffer_io_async: async io complete on unlocked page
appears occasionally. Not a good sign.

Error conditions are not handled too well. That is, if your disk gets corrupted, then it's possible for the kernel to hang. Note that the same caveat has always applied to the ext2 filesystem and people don't complain too much, but this is a problem it would be good to reduce or remove (in e2compr code, at least).

No other known bugs apply to the kernel. However, see sections Stability Comments and the TODO page of the manual.


There are no known bugs as such in e2fsprogs (i.e. e2fsck and others), but if your computer goes down while some files are being ext2-compressed, then e2fsck isn't as good at restoring things to a consistent state as it should be.

Philosophical bugs

The behaviour of lsattr with symbolic links is questionable. Also, ``not a typewriter'' is not a very appropriate error for files on non-ext2 filesystems.

The NOCOMPR flag should be scrapped; we can use an appropriate combination of O_NOTRANS and mandatory locking instead.

Anything else that you'd like to add, write to the e2compr mailing list or maintainer.

Stability Comments

e2compr has been around since Linux 1.2. Many people say they've been using e2compr for years, without any problems.

People's good experience should be balanced by what's written under the bugs section above, of course.

Some time ago, Peter did a couple of simple stress tests (looping around `make -j' in /usr/src/linux for a few hours while doing `cp -R' with a couple of directories, which is an attempt to catch race conditions) but no-one's done any serious (as in methodical, exhaustive) testing. Some other people using Antoine's patches for 1.2.13 or 1.3.56 commented that they'd had no problems. Antoine says that the patch for linux-1.2 did receive serious testing, but no-one else knows what the nature of that testing was.

Return to Main Entrance
e2compr maintainers
Last modified: Sun May 9 10:17:53 EST 1999